A revision of NASA SeaDAS atmospheric correction algorithm over turbid waters with artificial Neural Networks estimated remote-sensing reflectance in the near-infrared

Abstract

For atmospheric correction over turbid waters, due to non-negligible water-leaving radiance (Lw) in the near-infrared (NIR), measurements in the short-wave infrared (SWIR) are usually required to achieve reliable remote-sensing reflectance (Rrs). But several ocean color satellite sensors, such as the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) and other small satellites, have no bands in the SWIR domain. We here present an atmospheric correction algorithm (termed as ACANIR-NN) based on NASA SeaDAS (version 7.5.3), which can achieve atmospheric correction seamlessly over clear and turbid waters, even for sensors having no spectral bands in SWIR. Specifically, ACANIR-NN uses estimated Rrs(NIR) from available Rrs in the visible bands with a specifically designed artificial Neural Networks to carry out atmospheric correction, and the performance of ACANIR-NN is evaluated over eight coastal locations having ground measurements by the Aerosol Robotic Network-Ocean Color (AERONET-OC) system. It is found that the Mean Absolute Percent Difference (MAPD) of Rrs retrievals by ACANIR-NN for this dataset is smaller by a factor of two or more than that by the standard SeaDAS algorithm (termed as ACANIR-bio) for each band, especially for Rrs(412) and Rrs(443), which is 7.5% and 7.7%, respectively, from ACANIR-NN, but they are 44.0% and 27.5% from ACANIR-bio. We further demonstrated the applicability of ACANIR-NN to SeaWiFS measurements over turbid waters, where consistent Rrs products were also obtained compared to that generated from the same-day MODerate resolution Imaging Spectrometer (MODIS) measurements using SWIR bands. These results indicate that ACANIR-NN can generate reliable Rrs over turbid coastal areas, as well as clear ocean waters, for sensors having no SWIR bands.